Presentation given by Eva-Maria Egger at the ‘Migration and mobility – new frontiers for research and policy’ conference in Ghana, October 2017

1. The nature and impact of repeated
migration within households
in rural Ghana
Eva-Maria Egger (IFAD) and Julie Litchfield (University of
Sussex)
5th October 2017
UNU-WIDER conference in Accra, Ghana

2. Motivation
• Open question whether migration has positive or
negative impact on sending household  empirical
evidence needed
• Migration is a diverse phenomenon. People move for
many reasons (work, family, education) and
repeatedly and more than one family member might
leave.
• Within New Economics of Labour Migration (NELM),
but allowing for this diversity
• First migration specific panel study for Ghana

3. Research questions
• Are new migrants different from the previous
migrants of same household?
• How does having a new migrant affect the welfare
of households who already engage in migration?

4. Data
• Household panel 2013 and 2015 in five regions of
Ghana
• Collected by Migrating out of Poverty, Centre for
Migration Studies, University of Ghana
(supported by UKAID)
• Focus on migration:
– Oversample households with migrants
– Questionnaire covers migration history, remittances,
and return migrants

5. Survey regions
Survey regions
Other regions

6. Conceptual framework
Household member E is a ‘new’ migrant
Household member D is a returned migrant
Household with migration experience and ‘new’ migrant:

7. Description of new migrants and
their households
• New migrant households: larger, family farmers, more
of their migrants have job, more have returnee
• New migrants: younger generation, straight from
education or unpaid work, move for work, education,
marriage, few and low remittances, lower moving costs
• All migrants: permanent and migration is financed with
savings, i.e. credit constraint environment

8. Impact of new migrant on
welfare: Methodology
∆𝑌𝑖,𝑡 = 𝛽1 𝑁𝑒𝑤𝑀𝑖𝑔𝑖 + 𝛽2∆𝑋𝑖,𝑡 + 𝛽3∆𝐿𝑀𝑐,𝑡 + ∆𝜖𝑖,𝑡
• First difference model of wealth index (Y) on
indicator for new migrant (NewMig) and
observable household (X) and community
characteristics (LM)
• Endogeneity: Reverse causality and selection
– 1st difference takes care of time-invariant
unobservables
– Baseline entropy balancing weights reduce selection
by making households look comparable

9. Outcome variable: Asset
index
• Composite measure of housing quality (number of rooms,
presence of bathroom and toilet, wall material, floor material)
• Computed using Multiple Correspondence Analysis (similar to
Principle Component or Factor Analysis)

10. Results

12. Interpretation
• Asset index changes slowly and tends to rather
capture increase than decline
• Short period might also imply that positive effects
of remittance receipt haven’t materialised yet
• Low costs of new migrants’ move and low
remittances means no loss in labour
• Financing of migration through savings means
savings cannot be used for investments

13. Conclusion
• New panel study of migration in Ghana
• Repeated migration patterns and different motivations
for migration within the same household
• `New’ migrants often from younger generation, moving
relatively more for education and family reasons, pay
less for their move, remit rarely and less
• No impact found of having a new migrant on
households left-behind who already had engaged in
migration. Lower costs and use of savings can explain
result
• More longitudinal data and more outcome measures
needed for conclusive analysis

14. Appendix

15. Entropy balancing weights
• Ex-ante definition of balance:
• choose variables and moments (mean, variance…) to be balanced
• Compute weights and keep all observations that allow weights.
• Treated units have a weight of 1, control according to formula below.
• Run weighted least squares regression

16. Balance statistics

18. Community shocks